JP2008090372A - Storage device and load balancing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a storage device maintaining high-speed I/O throughput. <P>SOLUTION: The storage device includes: a load factor calculating part for calculating the load factors of a plurality of physical devices; a load factor determining part for determining whether or not the load factors of the physical devices calculated by the load factor calculating part are equal to or greater than a predetermined threshold value; a command ratio determining part for determining whether or not the command ratio of either a write command or a read command issued from a host computer is equal to or greater than a predetermined threshold value, if the load factor determining part determines that the load factors of the physical devices are equal to or greater than the predetermined threshold value; and a switching part for replacing data with parities among the plurality of physical devices if the command ratio determining part determines that the command ratio is equal to or greater than the predetermined threshold value. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a storage apparatus and a load distribution method, and is suitable for application to a storage apparatus that controls the load balance of physical disks, for example.

  In general, in order to bring out a high I / O processing performance in a storage apparatus composed of a plurality of disks, it is necessary to sufficiently draw out the performance of the disks constituting the back end.

  Here, many RAID (Redundant Arrays of Inexpensive Disks) configurations for improving the reliability of data have been adopted for the construction of the back end, and a storage apparatus capable of configuring various levels of RAID is proposed. (See, for example, Non-Patent Document 1).

In addition, as a method for realizing disk load balancing, when a specific logical volume set to a physical volume is overloaded, the contents of that logical volume are set to a spare physical volume. A method has been proposed in which the load is distributed by copying to a logical volume and providing both logical volumes as one virtual logical volume to the host device (see, for example, Patent Document 1).
Paper: "A Case for Redundant Arrays of Inexpensive Disks (RAID)" by David A. Patterson, Garth Gibson and Randy H. Katz [URL search on September 29, 2006] http://www.cs.cmu.edu /~garth/RAIDpaper/Patterson88.pdf JP 2006-053601 A

  Here, in a RAID configuration that performs striping control, data is distributed over a plurality of disks, and thus one process may be completed by accessing the plurality of disks. Therefore, if the load is concentrated on a specific disk that constitutes the RAID and the disk on which the load is concentrated reaches the processing limit, the disk becomes a bottleneck, causing a processing delay, and the storage device performance can be sufficiently extracted. There is a problem that it is not (hereinafter defined as a disk neck).

  Further, according to the method proposed in Patent Document 1, it is possible to distribute a load by allocating a spare physical volume disk and increasing the number of disks in charge for one logical volume. However, since a spare physical volume is prepared, there is a problem that the area that can be used by the user is reduced.

  The present invention has been made in view of the above points, and proposes a storage apparatus and a load distribution method that can maintain high-speed I / O processing performance and sufficiently draw out storage apparatus performance.

  In order to solve such a problem, in the present invention, a storage apparatus having a plurality of physical devices that distribute and hold data transmitted from a host computer and the parity of the data for each predetermined unit, A load factor calculation unit that calculates a load factor, a load factor determination unit that determines whether or not the load factor of the physical device calculated by the load factor calculation unit is equal to or greater than a predetermined threshold, and a physical device by the load factor determination unit Command ratio determination for determining whether the command ratio of either a write command or a read command issued from the host computer is greater than or equal to a predetermined threshold when it is determined that the load factor of the host is greater than or equal to a predetermined threshold And the command ratio determining unit determine that the command ratio is equal to or greater than a predetermined threshold and It was set to and a replacement unit for replacing the I.

  Therefore, when the load is concentrated on a specific physical device, in order to equalize the load, avoiding the situation where the physical device becomes a bottleneck and processing delay occurs and the storage device performance cannot be sufficiently extracted. Thus, it is possible to load balance physical devices.

  Further, in the present invention, there is provided a load distribution method for a storage apparatus having a plurality of physical devices that scatter-holds data transmitted from a host computer and the parity of the data for each predetermined unit. A first step of calculating a rate, a second step of determining whether or not the load factor of the physical device calculated in the first step is equal to or greater than a predetermined threshold, and a load of the physical device in the second step A third step of determining whether the command ratio of either a write command or a read command issued from the host computer is equal to or higher than a predetermined threshold when it is determined that the rate is equal to or higher than a predetermined threshold; When it is determined in the third step that the command ratio is equal to or greater than a predetermined threshold, data and parity between a plurality of physical devices And so and a fourth step of replacing.

  Therefore, when the load is concentrated on a specific physical device, in order to equalize the load, avoiding the situation where the physical device becomes a bottleneck and processing delay occurs and the storage device performance cannot be sufficiently extracted. Thus, it is possible to load balance physical devices.

  According to the present invention, the load factor of a plurality of physical devices is calculated, it is determined whether the load factor of the physical device is equal to or greater than a predetermined threshold, and the load factor of the physical device is equal to or greater than the predetermined threshold. When it is determined, it is determined whether the command ratio of either a write command or a read command issued from the host computer is equal to or greater than a predetermined threshold, and when it is determined that the command ratio is equal to or greater than the predetermined threshold When the load is concentrated on a specific physical device by exchanging data and parity between multiple physical devices, processing load occurs due to the physical device becoming a bottleneck in order to equalize the load. Therefore, it is possible to avoid the situation where the performance of the storage device cannot be sufficiently obtained and to load balance the physical device, thus, high-speed I / O processing performance Maintaining, can realize a storage apparatus and a load distribution method capable of eliciting a storage device performance sufficiently.

  Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a configuration diagram showing an overall configuration of a storage system 1 according to an embodiment of the present invention. In FIG. 1, the storage apparatus 10 includes a storage control apparatus 100 that controls the entire storage apparatus 10 and a storage drive apparatus 200 that provides a storage area.

  A host computer 300 such as an application server that is a host device that uses the storage apparatus 10 is connected to the storage apparatus 10 via a host adapter 110 of the storage control apparatus 100.

  As an interface for connecting the storage apparatus 10 and the host computer 300, for example, a SAN (Storage Area Network) 180 that is a dedicated network is used. The SAN 180 is constructed using Fiber Channel, SCSI (Small Computer System Interface), iSCSI (Internet Small Computer System Interface), USB (Universal Serial Bus), IEEE (Institute of Electrical and Electronic Engineers) 1394 bus, etc. Is possible. Further, a plurality of host computers 300 connected to the storage apparatus 10 may be used.

  Further, on the host computer 300, control software for operating and setting the disk configuration of the storage apparatus 10 and controlling the operation of the storage apparatus 10 is installed. The host computer 300 can control the operation of the storage apparatus 10 by issuing a command to the storage apparatus 10 using the control software on the host computer 300. However, the storage apparatus 10 and the storage control software on the host computer 300 exchange data via a LAN (Local Area Network) 190 and a NIC (Network Interface Card) 170 of the storage apparatus 10, for example. Note that a computer that manages and controls the storage apparatus 10 may be different from the host computer 300.

  The storage control device 100 includes a host adapter 110 that connects a host computer 300 that uses the storage device 10, a CPU (Central Processing Unit) 120 that performs overall control of the storage device 10, and the CPU 120 controls the storage device 10. Memory 130 for storing programs and data necessary for data transfer, cache memory 140 for temporarily storing data for transmission / reception with respect to the host computer 300, and providing a high-speed response to data access, parity when configuring RAID It includes ASICs (Application Specific Integrated Circuits) 150 that are application-specific integrated circuits that perform calculations and the like, and a disk adapter 160 that connects the storage control device 100 and the storage drive device 200.

  The memory 130 includes a program 1300 used by the CPU 120 to control the storage apparatus 10, an address management table 1310 used to manage data and parity storage location information in the physical disk 2100, and exchange of data and parity. A data / parity switching execution timing management table 1320 for managing information for determining an execution timing (hereinafter referred to as “data / parity switching”) is stored.

  The storage drive device 200 includes a large number of disk boxes 210, and the disk box 210 includes a large number of physical disks 2100 (storage devices). Here, for example, a plurality of physical disks 2100 form a RAID group 2200 as shown in FIG. 1, and a logical unit (LU) (not shown) that is a logical storage area in the RAID group 2200. Thus, a large-capacity storage area is provided to the host computer 300. In the present invention, the storage area is configured by RAID5 or RAID6. In this case, RAID 5 and RAID 6 are generally RAID levels in which data and the parity of the data are distributed and held in a plurality of physical disks 2100. The present invention can be applied not only to the physical disk 2100 but also to various other physical devices such as a semiconductor memory.

  FIG. 2 is a configuration diagram of the address management table 1310. The address management table 1310 includes a stripe column number column 1311, a logical address column 1312, and a physical address column 1313. Here, the stripe column indicates a control unit for holding corresponding data and parity at a RAID level in which data and parity such as RAID 5 and RAID 6 are distributed and held in a plurality of disks.

  The stripe column number column 1311 manages a number for uniquely identifying a stripe column in the RAID group 2200. The logical address column 1312 and the physical address column 1313 manage information on which address of the actual physical disk 2100 the logical address of the logical unit corresponds to. The physical address column 1313 includes a data address column 1314 for managing the address of the physical disk 2100 for storing data and a parity address column 1315 for managing the address of the physical disk 2100 for storing parity. FIG. 2 shows an example of 2D + 1P RAID 5 in which there are two physical disks 2100 for storing data and one physical disk 2100 for storing parity, and the parity address column 1315 in the same stripe column is 1 In the case of RAID 6, two stages of parity address storage columns 1315 in the same stripe column are created.

  The logical address column 1312 includes a LUN (Logical Unit Number) column 13121, a head LBA (Logical Brock Address) column 13122, and a tail LBA column 13123. The LUN column 13121 manages the numbers of logical units created in the RAID group 2200. The head LBA column 13122 and the tail LBA column 13123 manage the head and tail addresses of the logical address space of the logical unit.

  The data address column 1314 includes an HDD (physical disk) number column 13141, a head LBA column 13142, and a tail LBA column 13143. The parity address column 1315 includes an HDD number column 13151, a head LBA column 13152, and a tail LBA column 13153.

  The HDD number columns 13141 and 13151 manage the numbers of physical disks 2100 that are storage destinations of data and parity. The head LBA column 13142 corresponds to the head LBA column 13122, and the tail LBA column 13143 corresponds to the tail LBA column 13123. The logical address space recognized by the host computer 300 and the actual data storage destination, respectively. To manage the correspondence with the physical address space. The head LBA column 13152 and the tail LBA column 13153 manage the head and tail addresses of the physical address space where the parity is stored.

  In FIG. 2, physical location information of the physical disk 2100 is used as an example of the identifiers of the HDD number columns 13141 and 13151. For example, the HDD number “MN” indicates that the physical disk 2100 is mounted in the “N” column of the “M” -th disk box 210. Here, the “M” -th disk box 210 is the disk box 210 positioned at the “M” -th when the plurality of disk boxes 210 are arranged in a line. The address management table 1310 is created and managed for each RAID group 2200.

  FIG. 3 is a configuration diagram of the data / parity exchange execution timing management table 1320. The data / parity exchange execution timing management table 1320 includes a RAID group number column 1321, an HDD number column 1322, a physical disk load factor column 1323, an average physical disk load factor column 1324, an average write command rate column 1325, and an average read command rate column 1326. And a data / parity exchange execution attribute column 1327.

  The RAID group number column 1321 manages a number created in the storage drive apparatus 200 for uniquely identifying the RAID group 2200. The HDD number column 1322 manages numbers for identifying the respective physical disks 2100 constituting the RAID group 2200. The physical disk load factor column 1323 manages the current load status of the physical disk 2100, and is represented by, for example, the operation rate of the physical disk 2100.

  The average disk load factor column 1324 manages the average load factor of the physical disks 2100 belonging to each RAID group 2200 and is calculated using the physical disk load factor in the physical disk load factor column 1323. The average write command rate column 1325 manages the ratio of write commands issued to each RAID group 2200. The average read command rate column 1326 manages the ratio of read commands issued to each RAID group 2200.

  The data / parity exchange execution attribute column 1327 manages parameters referred to when issuing an execution command for a data / parity exchange process to be described later. When it is necessary to execute data / parity exchange processing, “read” (not shown in FIG. 3) or “write” is held, and when data / parity exchange execution is not necessary, “N / A” is held. To do. Also, “N / A” is held when the RAID group 2200 is newly created. Further, as a component of the data / parity exchange execution timing management table 1320, a command rate of a logical unit configured in the RAID group 2200 may be held.

  The data / parity exchange execution timing management table 1320 is updated each time the CPU 120 processes each command of the read command and the write command issued from the host computer 300 to the storage apparatus 10, and will be described later. At the time of execution, it is referred to as information for determining whether or not to execute.

  FIG. 4 is a flowchart of the data / parity exchange execution timing management table update process. When the CPU 120 receives an I / O command such as a read command or a write command issued from the host computer 300 to the storage apparatus 10, the CPU 120 starts updating the data / parity exchange execution timing management table 1320 and receives the received I / O command. I / O processing is executed (step S301).

  Thereafter, the CPU 120 calculates the physical disk load factor of the physical disk 2100 belonging to the RAID group 2200 with respect to the RAID group 2200 in which the disk access has occurred in accordance with the I / O processing in step S301, and the physical disk load factor. The column 1323 is updated (step S302).

  Next, the CPU 120 calculates the average load factor of the physical disks 2100 belonging to the RAID group 2200 based on the information in the disk load factor column 1323, and updates the average load factor column 1324 (step S303).

  Thereafter, the CPU 120 calculates the average write command rate issued to the RAID group 2200, and updates the average write command rate column 1325 (step S304). Thereafter, the CPU 120 calculates the average read command rate issued to the RAID group 2200, and updates the average read command rate column 1326 (step S306). The CPU 120 ends the update process of the data / parity exchange execution timing management table 1320 through the above steps. However, the update timing of the data / parity exchange execution timing management table 1320 may be executed before the start of the I / O process or during the I / O process.

  FIG. 5 is a flowchart of the data / parity switching process. The CPU 120 starts data / parity exchange processing at a predetermined timing (described later), and data / parity exchange is necessary for all RAID groups 2200 based on information in the data / parity exchange execution timing management table 1320. (Step S401) (details will be described with reference to FIG. 6). If the CPU 120 determines that there is a RAID group 2200 that requires data / parity switching (S401: YES), the CPU 120 executes data / parity switching for the target RAID group 2200 (step S402) (for details). Explained in FIG. Thereafter, the CPU 120 ends the data / parity switching process.

  On the other hand, when CPU 120 determines in step S401 that there is no RAID group 2200 that requires data / parity exchange execution (S401: NO), CPU 120 ends the data / parity exchange execution process as it is.

  The data / parity exchange process is executed at regular timing or at arbitrary timing. Examples of the periodic timing include execution at regular intervals by a timer, execution every time the number of received commands from the host computer 300 reaches a multiple of a specific number, and the like. As an example of arbitrary timing, when the administrator of the storage apparatus 10 gives an instruction to execute this processing using the control software installed in the host computer 300, or the storage apparatus 10 receives a specific command. Or the CPU 120 executes a specific command.

(1) Data / Parity Exchange Execution Necessity Determination Process FIG. 6 is a detailed flowchart of step S401 for executing the data / parity exchange execution necessity determination process. The CPU 120 determines, for a specific RAID group 2200, whether or not there is a physical disk 2100 with a high load factor in the RAID group 2200 based on the physical disk load factor column 1323 (step S501).

  If the CPU 120 determines that there is a disk 2100 with a high physical disk load factor (S501: YES), the CPU 120 determines whether or not the physical disk load factor varies among the physical disks 2100 in the RAID group 2200. Determination is made (step S502).

  Next, when the CPU 120 determines that the physical disk load factor varies among the physical disks 2100 in the RAID group 2200 (S502: YES), the write command issued to the RAID group 2200 is dominant. Is determined based on the average write command rate column 1325 (step S503). If the CPU 120 determines that the write command is dominant (S503: YES), the CPU 120 changes the data / parity exchange execution attribute 1327 to “write” (step S504), and the determination result in step S401 is data. / Proceed to “YES” indicating that parity replacement is required.

  On the other hand, when the CPU 120 determines that the write command is not dominant (S503: NO), the CPU 120 determines whether the read command issued to the RAID group 2200 is dominant or not. A determination is made based on the field 1326 (step S505). If the CPU 120 determines that the read command is dominant (S505: YES), it changes the data / parity exchange execution attribute 1327 to “read” (step S506), and the determination result in step S401 is data. / Proceed to “YES” indicating that parity replacement is required.

  On the other hand, when the CPU 120 determines that there is no disk 2100 with a high physical disk load factor (S501: NO), the CPU 120 determines that there is no variation in the physical disk load factor among the physical disks 2100 in the RAID group 2200 ( If it is determined that the read command is not dominant (S505: NO), the data / parity exchange execution attribute 1327 is changed to “N / A” (step S507), and the determination result of step S401 Advances to “NO” indicating that data / parity switching is not required.

  Here, when the disk load factor is a high value in step S501, for example, the disk load factor is 100% or 95% or more, which is a predetermined threshold value or more, and the physical disk 2100 performs processing. The case where the limit is approached or the processing limit is reached is shown.

  In addition, when the physical disk load factor varies in step S502, for example, when the average disk load factor 1324 of a specific RAID group 2200 is compared with the physical disk load factor of each physical disk 2100, the difference Is greater than or equal to a set threshold, or the absolute value of the difference between the maximum and minimum disk load factors is greater than or equal to the set threshold.

  Furthermore, when there is a dominant command in step S503, for example, when the setting threshold is set to 60%, the ratio of the issued write command to the read command is 80%: 20%. In this case, one of the command ratios is equal to or greater than the set threshold.

  Note that the setting threshold value used for the determination in step S502, the specific value of the dominant command ratio setting threshold value in step S503, and the like are changed as tuning parameters that can be set by the control software installed in the host computer 300. Is possible.

(2) Data / Parity Exchange Execution Process FIG. 7 is a detailed flowchart of step S402 for executing the data / parity exchange execution process. When the data / parity exchange execution necessity determination process in step S401 ends, the CPU 120 starts the data / parity exchange execution process, and determines whether the data / parity exchange execution attribute column 1327 of the RAID group 2200 is “write”. Is determined (step S601).

  When the data / parity exchange execution attribute column 1327 is “write” (S601: YES), the CPU 120 determines that the physical disk 2100 in the RAID group 2200 has a higher physical disk load factor than the physical disk 2100. A data / parity exchange execution command is issued to exchange data and parity so as to reduce the ratio of holding parity (step S602).

  On the other hand, when the data / parity exchange execution attribute column 1327 is not “write” (S601: NO), the CPU 120 determines that the data / parity exchange execution attribute column 1327 is “read”, and the RAID. A data / parity exchange execution command is issued to the physical disks 2100 of the group 2200 to exchange data and parity so as to increase the ratio of holding parity in the physical disks 2100 with a high physical disk load factor (step S603). .

  Then, the CPU 120 executes the issued exchange command to exchange data and parity (step S604). Then, the CPU 120 ends the data / parity switching execution process, and then ends the data / parity switching process.

  FIG. 8 is a detailed flowchart of step S604 of executing the data / parity exchange command execution processing. When the CPU 120 issues a data / parity exchange execution command in step S602 or step S603, the CPU 120 specifies the current storage position and the replacement storage position of the replacement target data and parity based on the information in the address management table 1310. (Step S701).

  Next, the CPU 120 reads the replacement target data and parity from the physical disk 2100 to the cache memory 140 (step S702). Subsequently, the CPU 120 transfers the replacement target data and parity read out to the cache memory 140 to the new address that is the storage location of the replacement destination (step S703). Eventually, the CPU 120 updates (replaces) the physical disk 2100 by updating the address values of the HDD number 13141, the head LBA 13142 and the tail LBA 13143, and the parity HDD number 13151, the head LBA 13152, and the tail LBA 13153 of the data to be replaced. The address management table 1310 for managing the storage location is updated (step S704). Thereafter, the CPU 120 ends the data / parity exchange command execution process.

  In this case, when the data / parity exchange execution attribute column 1327 is “write”, the CPU 120 holds the parity in the physical disk 2100 having a low physical disk load factor with respect to the physical disk 2100 of the RAID group 2200. Data and parity are interchanged so as to increase the ratio. On the other hand, when the data / parity exchange execution attribute column 1327 is “read”, the CPU 120 holds the parity of the physical disk 2100 having a high physical disk load ratio with respect to the physical disk 2100 of the RAID group 2200. The data and the parity are switched so as to increase the value.

  Here, an example of replacement of data and parity will be described with reference to FIGS. FIG. 9 and FIG. 10 show an example of replacement in RAID 5 composed of three physical disks. FIG. 9 shows each data storage structure in the physical disk 2100 at the normal time (data / parity switching is not performed). In FIG. 9, for example, data is stored such that the parity ratio between the three physical disks 2100 is 1: 1: 1.

  At this time, in general, in RAID 5 or RAID 6, when performing data update (write) processing, the physical disk 2100 that holds the parity tends to be more heavily loaded than the physical disk 2100 that holds the data. This is because when a data update occurs, an old data read and an old parity read for creating a new parity, and a new data write and a new parity write for writing new data and a new parity to the disk are respectively performed on the respective physical disks 2100. However, since data is held in a plurality of physical disks 2100, access to all physical disks 2100 does not necessarily occur depending on the update range, whereas parity is held in one physical disk 2100. This is because access is always generated every time data is updated, and thus a difference in access frequency occurs between data and parity. Therefore, in an I / O environment in which data write is dominant, for example, when the load is concentrated only on a specific physical disk 2100 of the RAID group 2200 configured with RAID 5, the load is concentrated and is in a high load state. By making the parity ratio of the physical disk 2100 lower than that of the other physical disk 2100 and increasing the parity ratio of the other physical disk 2100 with low load than that of the physical disk 2100 in the high load state, Can be distributed.

  On the other hand, in the read process, since no access to the parity occurs, the physical disk 2100 that holds the parity is in a lower load state than the physical disk 2100 that holds the data. Therefore, in an I / O environment in which data read is dominant, the parity ratio of the physical disk 2100 in a high load state is set higher than that of the other physical disks 2100, and the parity ratio of the other physical disk 2100 having a low load is set. Is lower than that of the physical disk 2100 in a high load state, load distribution among the physical disks 2100 becomes possible.

  Specifically, in the storage apparatus 10, it is determined that the physical disk load ratio of the physical disk 2100-3 of the RAID group 2200 shown in FIG. 9 is high, and the physical disk load ratio varies among the physical disks 2100 in the RAID group 2200. If the write command issued to the RAID group 2200 is determined to be dominant, the parity holding ratio in the physical disk 2100-3 having a high physical disk load factor is lowered. Thus, the data and the parity are switched. That is, in the storage apparatus 10, the parity of the physical disk 2100-3 in the corresponding stripe column and the data of the physical disk 2100-1 or 2100-2 are exchanged.

  Further, in the storage apparatus 10, it is determined that the physical disk load factor of the physical disk 2100-3 in the RAID group 2200 shown in FIG. 9 is high, and the physical disk load factor varies among the physical disks 2100 in the RAID group 2200. When the read command issued to the RAID group 2200 is determined to be dominant, the parity holding ratio in the physical disk 2100-3 having a high physical disk load factor is increased. Swap data and parity. That is, in the storage apparatus 10, the data of the physical disk 2100-3 in the corresponding stripe row and the parity of the physical disk 2100-1 or 2100-2 are exchanged.

  FIG. 10 shows the data storage structure in the physical disk 2100 after the data / parity exchange is performed. FIG. 10 shows a state in which, for example, the positions of data and parity in the same stripe column are respectively switched, and the parity ratio between the three physical disks 2100 is changed to 1: 1: 2. That is, it is determined that the physical disk load rate of the physical disk 2100-3 of the RAID group 2200 shown in FIG. 9 is high, the physical disk load rate is determined to vary among the physical disks 2100 in the RAID group 2200, and The state after the change of the parity retention ratio when it is determined that the read command issued to the RAID group 2200 is dominant is shown.

  In addition, the replacement pattern for designating the parity ratio of the physical disk 2100 is the tuning parameter that can be changed by the administrator of the storage apparatus 10 by the control software installed in the host computer 300 by the administrator. It can be changed. Furthermore, as a replacement pattern setting item, a target range in which the positions of data and parity are switched in one process may be specified. In this case, the target range includes all the physical disk areas at once, or an arbitrary stripe column unit.

  In this way, in the storage system 1, when the load is concentrated on a specific physical disk 2100 in the RAID group 2200 and the read command or write command is dominant, the parity of the physical disk 2100 in the RAID group 2200 is determined. By changing the ratio, data and parity are switched for each stripe column of the physical disk 2100 in the RAID group 2200.

  Therefore, when the load is concentrated on a specific physical disk 2100, the load is evenly distributed among the physical disks 2100 in the same RAID group 2200. Therefore, the physical disk 2100 can be load-balanced while avoiding a disk neck. Thus, high-speed I / O processing performance can be maintained.

  In other words, in the present invention, when the load is concentrated only on a specific physical disk, the parity ratio of the physical disks constituting the RAID group is changed. Therefore, it is possible to realize a storage apparatus that can equalize the load balance of physical disks, avoid the occurrence of a disk neck, and maintain high-speed I / O processing.

  The present invention can be widely applied to various storage apparatuses other than a RAID group configured from a plurality of physical disks.

It is a block diagram which shows the schematic structure of the storage system by this Embodiment. It is a conceptual diagram with which it uses for description of an address management table. It is a conceptual diagram with which it uses for description of a data / parity exchange execution opportunity management table. It is a flowchart with which it uses for description of a data / parity exchange execution opportunity management table update process procedure. It is a flowchart with which it uses for description of a data / parity interchange process procedure. It is a flowchart with which it uses for description of a data / parity exchange execution necessity judgment processing procedure. It is a flowchart with which it uses for description of a data / parity exchange execution processing procedure. It is a flowchart with which it uses for description of a data / parity exchange command execution processing procedure. It is a conceptual diagram with which it uses for description of the example of data / parity interchange. It is a conceptual diagram with which it uses for description of the example of data / parity interchange.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Storage system, 10 ... Storage apparatus, 120 ... CPU, 300 ... Host computer, 1310 ... Address management table, 1320 ... Data / parity exchange execution timing management table, 1323 ... Physical disk load factor column 1324 ... Average physical disk load factor column, 1325 ... Average write command rate column, 1326 ... Average read command rate column, 1327 ... Data / parity exchange execution attribute, 2100 ... Physical disk, 2200 ... RAID group

Claims (16)

A storage device having a plurality of physical devices that distribute and hold data transmitted from a host computer and the parity of the data for each predetermined unit,
A load factor calculation unit for calculating a load factor of the plurality of physical devices;
A load factor determining unit that determines whether the load factor of the physical device calculated by the load factor calculating unit is equal to or greater than a predetermined threshold;
When the load factor determining unit determines that the load factor of the physical device is equal to or greater than a predetermined threshold, the command ratio of either a write command or a read command issued from the host computer is equal to or greater than a predetermined threshold. A command ratio determining unit for determining whether or not there is,
A storage apparatus comprising: a replacement unit that replaces the data and the parity between the plurality of physical devices when the command ratio determination unit determines that the command ratio is equal to or greater than a predetermined threshold. . The load factor determination unit
For a group composed of a predetermined number of physical devices among the plurality of physical devices, determine whether the load factor of any of the physical devices in the group is equal to or greater than a predetermined threshold value,
The command ratio determining unit
Determining whether the command ratio of either a write command or a read command issued from the host computer to the group is greater than or equal to a predetermined threshold;
The replacement unit is
The storage apparatus according to claim 1, wherein the data and the parity are exchanged between the physical devices in the group. The load factor calculation unit
Calculating an average load factor of the physical devices in the group;
The load factor determination unit
Determining whether or not a deviation between an average load factor of the physical devices in the group calculated by the load factor calculator and a load factor of the physical disks in the group is equal to or greater than a predetermined threshold;
The command ratio determining unit
When the load factor determination unit determines that the difference between the average load factor of the physical devices in the group and the load factor of the physical disks in the group is equal to or greater than a predetermined threshold, the command ratio is It is judged whether it is more than a predetermined threshold. The storage device according to claim 2 characterized by things. The load factor determination unit
Determining whether the absolute value of the difference between the maximum value and the minimum value of the load factor of the physical disk in the group is greater than or equal to a predetermined threshold;
The command ratio determining unit
When the load factor determination unit determines that the absolute value of the difference between the maximum value and the minimum value of the load factor of the physical disk in the group is equal to or greater than a predetermined threshold, the command ratio is equal to or greater than a predetermined threshold. The storage apparatus according to claim 2, wherein it is determined whether or not. The replacement unit is
If the command ratio determining unit determines that the command ratio of the write command is equal to or greater than a predetermined threshold, the parity of the physical device in the group determined by the load factor determining unit to be equal to or greater than the predetermined threshold is calculated. The storage apparatus according to claim 2, wherein data and parity are switched so as to reduce a holding ratio. The replacement unit is
6. The parity of a physical device in the group determined to be equal to or greater than a predetermined threshold by the load factor determination unit and data of another physical device in the group other than the physical device are exchanged. The storage device described in 1. The replacement unit is
When the command ratio determining unit determines that the command ratio of the read command is equal to or greater than a predetermined threshold, the parity of the physical device in the group determined by the load rate determining unit to be equal to or greater than the predetermined threshold is determined. The storage apparatus according to claim 2, wherein the data and the parity are switched so as to increase a holding ratio. The replacement unit is
The data of physical devices in the group determined to be equal to or greater than a predetermined threshold by the load factor determination unit and the parity of other physical devices in the group other than the physical device are switched. The storage device described in 1. A load distribution method for a storage apparatus having a plurality of physical devices that distribute and hold data transmitted from a host computer and the parity of the data for each predetermined unit,
A first step of calculating a load factor of the plurality of physical devices;
A second step of determining whether the load factor of the physical device calculated in the first step is equal to or greater than a predetermined threshold;
When it is determined in the second step that the load factor of the physical device is equal to or greater than a predetermined threshold, the command ratio of either a write command or a read command issued from the host computer is equal to or greater than a predetermined threshold. A third step of determining whether or not
And a fourth step of replacing the data and the parity between the plurality of physical devices when it is determined in the third step that the command ratio is equal to or greater than a predetermined threshold. Distribution method. In the second step,
For a group composed of a predetermined number of physical devices among the plurality of physical devices, determine whether the load factor of any of the physical devices in the group is equal to or greater than a predetermined threshold value,
In the third step,
Determining whether the command ratio of either a write command or a read command issued from the host computer to the group is greater than or equal to a predetermined threshold;
In the fourth step,
The load distribution method according to claim 9, wherein the data and the parity are exchanged between the physical devices in the group. In the first step,
Calculating an average load factor of the physical devices in the group;
In the second step,
Determining whether the difference between the average load factor of the physical devices in the group calculated in the first step and the load factor of the physical disks in the group is equal to or greater than a predetermined threshold;
In the third step,
When the load factor determination unit determines that the difference between the average load factor of the physical devices in the group and the load factor of the physical disks in the group is equal to or greater than a predetermined threshold, the command ratio is It is judged whether it is more than a predetermined threshold. The load distribution method according to claim 10 characterized by things. In the second step,
Determining whether the absolute value of the difference between the maximum value and the minimum value of the load factor of the physical disk in the group is greater than or equal to a predetermined threshold;
In the third step,
When it is determined in the second step that the absolute value of the difference between the maximum value and the minimum value of the load factor of the physical disk in the group is equal to or greater than a predetermined threshold, the command ratio is equal to or greater than the predetermined threshold. It is judged whether there exists. The load distribution method of Claim 10 characterized by the above-mentioned. In the fourth step,
If it is determined in the third step that the command ratio of the write command is greater than or equal to a predetermined threshold, the parity of the physical device in the group that is determined to be greater than or equal to the predetermined threshold in the second step is retained. The load distribution method according to claim 10, wherein data and parity are switched so as to reduce the ratio. In the fourth step,
The parity of the physical device in the group determined to be greater than or equal to a predetermined threshold in the second step and the data of another physical device in the group other than the physical device are replaced with each other. The load balancing method described. In the fourth step,
If it is determined in the third step that the command ratio of the read command is greater than or equal to a predetermined threshold, the parity of the physical device in the group that is determined to be greater than or equal to the predetermined threshold in the second step is retained. The load balancing method according to claim 10, wherein data and parity are switched so as to increase the ratio. In the fourth step,
The data of the physical device in the group determined to be equal to or greater than a predetermined threshold in the second step and the parity of another physical device in the group other than the physical device are exchanged. The load balancing method described.

Images